EP4208725B1 - Estimateur d'état de santé de batterie - Google Patents

Estimateur d'état de santé de batterie Download PDF

Info

Publication number
EP4208725B1
EP4208725B1 EP21766146.1A EP21766146A EP4208725B1 EP 4208725 B1 EP4208725 B1 EP 4208725B1 EP 21766146 A EP21766146 A EP 21766146A EP 4208725 B1 EP4208725 B1 EP 4208725B1
Authority
EP
European Patent Office
Prior art keywords
battery
dnn
supervised
unsupervised
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
EP21766146.1A
Other languages
German (de)
English (en)
Other versions
EP4208725A1 (fr
Inventor
Melanie SENN
Joerg Christian WOLF
Lorenz HAGHENBECK EMDE
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Audi AG
Volkswagen AG
Original Assignee
Audi AG
Volkswagen AG
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Audi AG, Volkswagen AG filed Critical Audi AG
Publication of EP4208725A1 publication Critical patent/EP4208725A1/fr
Application granted granted Critical
Publication of EP4208725B1 publication Critical patent/EP4208725B1/fr
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/392—Determining battery ageing or deterioration, e.g. state of health
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
    • G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/088—Non-supervised learning, e.g. competitive learning

Definitions

  • the present disclosure relates to methods, apparatuses, and systems for a battery state of health estimator and, more particularly, to a battery state of health estimator using induced domain knowledge for supervised learning.
  • Deep neural networks are widely used for many artificial intelligence (“AI”) applications including computer vision, autonomous vehicles, speech recognition, language translations, advertising, cancer detection, and robotics.
  • Machine learning methods to develop DNNs are divided into supervised learning and unsupervised learning.
  • supervised learning trustworthy labels are given as output data (i.e., the ground truth) in addition to the given input data to train a DNN.
  • unsupervised learning labels are not available and/or not trusted since the accuracy of such measurements are insufficient. Rather, the outputs are grouped together in clusters based on their similarity in the feature space.
  • SoH state of health
  • the SoH can be generally estimated by experimental and model-based methods.
  • Experimental methods monitor battery behaviors by analyzing the full cycle of experimental data of battery voltage, current, and temperature.
  • Model-based estimation methods can be further divided into physical mechanism-based state estimation methods and data-driven methods.
  • SoC status of charge
  • Figure 1 illustrates a prior art block diagram for determining battery SoH using an unsupervised estimator.
  • Unsupervised learning can be implemented by clustering (e.g. k-means clustering) on features from an input signal (e.g., observed battery characteristics) or by training an autoencoder as an unsupervised DNN 10.
  • cluster identification the goal is to minimize the similarity in data between clusters and at the same time maximize the similarity in the data within clusters. If adequate features are extracted from input data, desired cluster borders indicating battery state of health can be found, which shows clearly separable clusters. However, undesired cluster borders are found since it is extremely difficult to choose features properly.
  • the clustering algorithm is not suited for the available data since parameter thresholds are poorly formulated. The result is a mix of overlapping features in one cluster to another cluster, thereby obfuscating any useful information.
  • Prior art SoH estimation methods using neural networks are known from US2019/019080 and CN111027625 .
  • the invention provides a computing system according to claim 1, a corresponding method according to claim 8, and a corresponding data carrier according to claim 15.
  • Physics-based phenomena can be modeled using either domain-knowledge driven physical models that describe a limited number of identified dominating mechanisms (such as thermal, electrochemical, and aging effects in estimation of battery life) or purely data-driven models such as Kalman filters or machine learning.
  • various approaches include: (1) physical and chemical modelling including battery aging effects, (2) a weighted approach based on outlier detection, and (3) SN curves as an event-oriented concept.
  • a data-driven approach can be used for prediction of battery life taking into account any capacity degradation. For instance, a battery's status of charge is estimated by Kalman filters, while SoH, including degradation due to aging, can be determined by least squares regression. Inducing knowledge about the physical behavior can improve model accuracy by a physical-driven feature extraction.
  • both modeling from domain knowledge and data can be combined to minimize any remaining error from an analytical model that describes only a limited number of major mechanisms.
  • modelling remaining error of an analytical model with machine learning allows for a decrease in the overall modeling error (and consequently an increase in the model accuracy).
  • the remaining error is treated from a black box perspective and hence does not allow for identifying influences of new mechanisms.
  • induction of domain knowledge into machine learning allows for deriving some basic domain knowledge from a design of experiments table, and enables the identification of new mechanisms by analyzing features governed by domain knowledge (e.g. more electrolyte degradation in lithium-ion batteries that experience high temperature gradients).
  • the design of experiments table can include operating conditions of the experiments, such as outside temperature of the vehicle with a battery under test, set temperature of the controlled experiment, humidity of the controlled experiment, charging rate of the battery under test, a driver type for the corresponding battery under test (e.g., aggressive driver, moderate driver, or gentle driver relative to usage of the battery under test), etc.
  • the present disclosure may include domain knowledge from application engineering and machine learning from a data-centered approach.
  • domain knowledge is neglected in purely data-driven models that only apply an optimization algorithm to extract optimal features from input data.
  • domain knowledge can be used to derive characteristic features by a machine learning method (e.g., linear regression, logistic regression, etc.).
  • domain knowledge is induced into supervised learning for a DNN to determine a battery SoH estimation.
  • machine-learning models are optimized for deployment in vehicles with limited compute and memory capabilities.
  • a vehicle can represent battery degradation in SoC and/or SoH in real time given computational constraints.
  • the present disclosure uses a supervised learning approach to induce knowledge from design of experiments and transfer that induced knowledge from cell, to module, and finally to pack level in the vehicle for real world driving and charging scenarios.
  • Improved predictions are achieved by using domain knowledge as trustworthy labels in a supervised DNN to extract better features for an unsupervised DNN in order to benefit from both their strengths.
  • Figure 2 illustrates a block diagram for using process parameters and observed battery characteristics to generate a state of health indication for a battery.
  • An underlying dataset contains measurable battery characteristics (such as a time series of temperature, voltage, and current signals for a battery) resulting from process parameters such as set temperature, SoC at start and at end of the experiment, and charging and discharging rates for the experiments. If reliable labels are not available to quantify the battery state (such as measurements of battery capacity, battery SoH, and/or internal resistance of the battery), a supervised approach can be introduced.
  • a supervised classification DNN is trained by inducing the domain knowledge of the process parameters 20 as class labels.
  • the DNN's features are used for clustering.
  • Different class labels can be obtained by binning (e.g., dividing continuous values into discretized bins, such as 40 - 55 degrees Celsius for high temperatures) the values of the process parameters (e.g. low / medium / high C-rate or temperature).
  • a regression DNN can be trained to approximate the process parameters as output quantities directly without binning. In either cases, the features in the convolutional layers characterize the input data of observable battery characteristics 22 and the underlying process parameters 20.
  • the features are enforced to develop during training with respect to input quantities and process parameters at the same time.
  • the features are used for unsupervised training of an artificial intelligence ("AI") estimator.
  • AI artificial intelligence
  • Supervised training can be performed on the AI estimator to develop a supervised DNN for outputting state of health indications 24 based on observable battery characteristics 22.
  • the AI estimator can be an AI algorithm (e.g., a machine learning algorithm, deep learning algorithm, or combination thereof) trained using unlabeled data (i.e., without ground truth prediction reference from measurements or simulations).
  • Figure 3 illustrates a flow chart for an embodiment of the present disclosure to generate a state of health indication for a battery.
  • battery characteristics and process parameters for the battery are received to be used as training data 40.
  • the battery characteristics include observable quantities of the battery.
  • a BMS may be coupled to the battery for management of the battery load and cooling of the battery.
  • the BMS may measure various characteristics of the battery, including a battery temperature, a battery voltage, a battery electrical current, number of charge cycles, charging profile, discharging profile, etc. The measured characteristics can be measured over a predefined period time to generate a function of the specific characteristics.
  • the process parameters are generated by measuring the external factors applied to the battery, e.g., set temperature of the design of experiments, the start and/or end of the SoC, charge rate, etc.
  • the process parameters can individually influence a particular battery's workload and can lead to an individualized degradation for that particular.
  • a supervised deep neural network (“DNN”), referred to as a first supervised DNN, is trained using the received characteristics as input and received process parameters as outputs 42.
  • the supervised DNN can be trained using various supervised learning algorithms, including convolutional layers (plus max pooling) with a couple of fully connected layers, ResNet, U-Net, or other DNN architectures. In this manner, the features of the supervised DNN are induced from the known quantities of the received battery characteristics and the process parameters.
  • the features from the supervised DNN are extracted and then used for training an unsupervised AI estimator 44.
  • the unsupervised AI estimator can be trained based on the extracted features from the supervised DNN using one or more clustering methods. For instance, a machine learning algorithm for clustering such as k-means can be applied.
  • a machine learning algorithm for clustering such as k-means can be applied.
  • the received battery characteristics are inputted to the unsupervised AI estimator. Identified clusters can be found from the AI estimator.
  • another step can be included in which at least one flatten layer is removed from the supervised DNN so that the features can be extracted and remapped for another class outputs (e.g., SoH indications).
  • a second supervised DNN is trained using identified clusters (also referred to as "groups") from the unsupervised AI estimator 46.
  • the outputs of the unsupervised AI estimator can be clustered using a classification algorithm. For instance, outputs of the unsupervised DNN can be clustered by dividing continuous values into discretized bins. Alternatively, the outputs can be clustered using a DNN regression algorithm. It can be appreciated that other classification/ clustering algorithms can be used as well. Once clustered, the clustered outputs are validated with SoH indications 48.
  • battery characteristics (or also referred to as user battery data) for a specific battery are inputted to the second supervised DNN to determine a SoH indication 50.
  • High-effort SoH measurements can be performed for representative, measurable battery characteristics to assign clusters of the output of the second supervised DNN to a SoH indication (e.g. low / medium / high health or 10 / 50 / 90% battery health).
  • the user battery data can be the received battery characteristics, a battery data set for a specific battery, or a combination thereof.
  • user battery data is monitored and stored by a BMS, e.g., battery usage, charging and discharging cycle data for the battery, etc., giving a user profile corresponding to that battery.
  • a BMS e.g., battery usage, charging and discharging cycle data for the battery, etc.
  • the second supervised DNN can be retrained using the validated SoH indications to further refine the feature map of the second supervised DNN (e.g. to customize the estimator to different regions / drivers).
  • Figure 4 illustrates a flow chart for another embodiment of the present disclosure for generating a state of health indication for a battery.
  • the overall process of inducing domain knowledge for SoH estimation can include combining unsupervised (e.g., clustering) and supervised (e.g., classification/regression) training approaches.
  • the training of the unsupervised AI estimator 44b and training of the second supervised DNN 46 can be repeated a number of times until convergence condition is met 47. For instance, the steps 44b and 46 can be repeated based on a predefined number of repetitions, modified clustering convergence criteria, whether a predefined accuracy is achieved by training the second DNN, or combination thereof.
  • the features are extracted from the previous, unsupervised AI estimator and used for retraining.
  • outputs of the unsupervised AI estimator can be refined from a coarse number to a finer number of classifications.
  • Figure 5 illustrates a diagram for applying supervised learning to a deep neural network to determine known process parameters for a battery.
  • a supervised DNN 42 comprises convolutional layers 60 for feature extraction, a flatten layer 62, fully-connected layer(s) 64, and an output layer 66.
  • the flatten layer 62 is between the convolutional layers 60 for feature extraction and the fully-connected layer(s) for regression.
  • the output layer 66 provides classifications or regression values based on the output of the fully-connected layer(s) 64.
  • the features from the convolution layers are extracted as vectorized quantity (with flatten layer) for feature-based clustering.
  • This feature vector can then be used for generating clusters as it relates to SoH indications.
  • the identified clusters in the feature vector space may have related characteristics for representation as desired clusters for battery SoH estimation.
  • the identified clusters can then be relabeled with respect to the correlated, but unlabeled output of SoH values.
  • SoH values can include a percentage relative to a total ideal capacity of 100% for a non-degraded battery. For instance, 10% may mean that the battery under test may have degraded to the point where it can only hold preset percentage of the charge of a non-degraded battery.
  • the SoH indications can comprise relative terminology of various states, including "bad", “medium”, or "good” SoH indications. The various states can correspond to threshold percentage values of the SoH or SoC.
  • Figure 6 illustrates a diagram for applying unsupervised learning to a deep neural network.
  • Observable battery characteristics are inputted to convolutional layers 78.
  • the convolutional layers are extracted features from a supervised DNN that was trained with the battery characteristics and the process parameters.
  • the convolutional layers output to a flatten layer 80 to generate feature vectors 82.
  • the flatten layer transforms the 2D and/or 3D features from convolutional layers 78 to vector quantities.
  • the feature vectors 82 from the output of the flatten layer 80 can be clustered using one or more clustering algorithms to generate borders around similar feature vectors.
  • Figure 7 illustrates a graph diagram for clustering outputs of an unsupervised deep neural network.
  • the feature vectors 82 can be clustered using an unsupervised learning algorithm. Since the feature extraction from the supervised DNN (including process parameters as domain knowledge) is used to generate the feature vectors 82, the borders of the feature vectors 82 are more easily determined based on more meaningful features.
  • the clustering algorithm may identify cluster 1, cluster 2, cluster 3, ..., cluster N-1, and cluster N as distinct clusters.
  • the number of clusters can be selected based on a predefined amount or dependent on a specific threshold.
  • Figure 8 illustrates a graph diagram for validating SoH classifiers for unsupervised clusters in a feature vector space.
  • cluster 1 can be associated with a "Good” SoH indication.
  • representative sample(s) from each of the other clusters can be tested for a SoH indication and then the cluster can be associated with that SoH indication.
  • cluster 2 is associated with a "Bad” SoH indication;
  • cluster 3 is associated with "Medium” SoH indication.
  • FIG. 9 illustrates a diagram for a supervised DNN of the present disclosure for determining a battery SoH indication based on observable battery characteristics.
  • a supervised DNN 100 comprises convolutional layers 102, flatten layer 104, fully-connected layer(s) 106, and an output layer 108.
  • the output layer 108 can have SoH indications, e.g., "good", "bad", and "medium”.
  • the features of the supervised DNN can be derived from previous supervised (classification / regression by process parameters) and unsupervised (clustering) steps.
  • the unsupervised step was reformulated as a supervised step with the identified clusters as labels.
  • the supervised DNN 100 can be used to evaluate a particular battery for a SoH indication.
  • the observable battery characteristics are inputted to the second supervised DNN to generate a SoH indication of "good", "bad", and/or "medium”.
  • a vehicle 158 comprises a computing system 160, sensors 162, a vehicle communications system 164, a propulsion system 166, a control system 168, a power supply 170, a user interface system 172, and a battery management system 174.
  • the vehicle 158 may include more, fewer, and/or different systems, and each system may include more, fewer, and/or different components. Additionally, the systems and/or components may be combined and/or divided in a number of arrangements.
  • the computing system 160 may be configured to transmit data to, receive data from, interact with, and/or control one or more of the propulsion system 166, the sensors 162, the control system 168, BMS 174, and any other components of the vehicle 158.
  • the computing system 160 may be communicatively linked to one or more of the sensors 162, vehicle communications system 164, propulsion system 166, control system 168, power supply 170, user interface system 172, and BMS 174 by a system bus (e.g., CAN bus, Flexray, etc.), a network (e.g., via a vehicle-to-vehicle, vehicle-to-infrastructure, vehicle-to-device, and so on), and/or other connection mechanism.
  • a system bus e.g., CAN bus, Flexray, etc.
  • a network e.g., via a vehicle-to-vehicle, vehicle-to-infrastructure, vehicle-to-device, and so on
  • other connection mechanism
  • the computing system 160 may be configured to store data in a local data storage and/or communicatively coupled to an external data storage. It can be appreciated that the data can also be transmitted to a cloud service and received from the cloud service via over-the-air (“OTA") wireless techniques. For instance, OTA wireless technique can be used to transmit updated DNN models or to upload interesting data such as corner cases.
  • OTA wireless technique can be used to transmit updated DNN models or to upload interesting data such as corner cases.
  • the computing system 160 may be configured to cause the sensors 162 to capture images of the surrounding environment of the vehicle 158. In yet another embodiment, the computing system 160 may control operation of the propulsion system 166 to autonomously or semi-autonomously operate the vehicle 158. As yet another example, the computing system 160 may be configured to store and execute instructions corresponding to an algorithm (e.g., for steering, braking, and/or throttling) from the control system 170. As still another example, the computing system 160 may be configured to store and execute instructions for determining the environment around the vehicle 158 using the sensors 162. Even more so, the computing system 160 may be configured to store and execute instructions for operating the BMS 174 and a supervised DNN for indicating the SoH of a battery of the vehicle 158.
  • an algorithm e.g., for steering, braking, and/or throttling
  • the BMS 174 and/or computing system 160 can sense and record battery characteristics including, temperature, voltage, and current, via the CAN data every predefined number of hours and/or days.
  • the supervised DNN can use this information as input and generate a SoH indication for that battery. Based on the generated SoH indication, user feedback can be provided the user of the vehicle valuable direction to preserve the SoH of the battery (e.g., do not perform three fast charges in a row).
  • the computing system 160 can include one or more processors. Furthermore, the computing system can have its own data storage and/or use an external data storage.
  • the one or more processors may comprise one or more general-purpose processors and/or one or more special-purpose processors. To the extent the processor includes more than one processor, such processors could work separately or in combination.
  • Data storage of the computing system 160 may comprise one or more volatile and/or one or more non-volatile storage components, such as optical, magnetic, and/or organic storage.
  • the data storage may be integrated in whole or in part with the one or more processors of the computing system 160 and may be communicatively coupled to the data storage.
  • data storage of the computing system 160 may contain instructions (e.g., program logic) executable by the processor of the computing system 160 to execute various vehicle functions (e.g., the methods disclosed herein).
  • the term computing system may refer to data processing hardware, e.g., a CPU and/or GPU, and encompasses all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, multiple processors, computers, cloud computing, and/or embedded low-power devices (e.g., Nvidia Drive PX2).
  • the system can also be, or further include, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).
  • the system can optionally include, in addition to hardware, code that creates an execution environment for computer programs, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, or a combination of one or more of them.
  • a computer program can also be used to emulate the computing system.
  • a computer program which may also be referred to or described as a program, (software, a software application, an app, a module, a software module, a script, or code) can be written in any form of programming language, including compiled or interpreted languages, or declarative or procedural languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.
  • a computer program may, but need not, correspond to a file in a file system.
  • a program can be stored in a portion of a file that holds other programs or data, e.g., one or more scripts stored in a markup language document, in a single file dedicated to the program in question, or in multiple coordinated files, e.g., files that store one or more modules, sub-programs, or portions of code.
  • a computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a data communication network.
  • the processes and logic flows described in this specification can be performed by one or more programmable computers executing one or more computer programs to perform functions by operating on input data and generating outputs.
  • the processes and logic flows can also be performed by special purpose logic circuitry, e.g., an FPGA or an ASIC, or by a combination of special purpose logic circuitry and one or more programmed computers.
  • Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including by way of example semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks.
  • semiconductor memory devices e.g., EPROM, EEPROM, and flash memory devices
  • magnetic disks e.g., internal hard disks or removable disks
  • magneto-optical disks e.g., CD-ROM and DVD-ROM disks.
  • Embodiments of the subject matter described in this specification can be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a client computer having a graphical user interface, a web browser, or an app through which a user can interact with an implementation of the subject matter described in this specification, or any combination of one or more such back-end, middleware, or front-end components.
  • the components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include wired and/or wireless local area networks (“LANs”) and wired and/or wireless wide area networks (“WANs”), e.g., the Internet.
  • LANs local area networks
  • WANs wireless wide area networks
  • the computing system can include clients and servers.
  • a client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
  • a server transmits data, e.g., an HTML page, to a user device, e.g., for purposes of displaying data to and receiving user input from a user interacting with the device, which acts as a client.
  • Data generated at the user device e.g., a result of the user interaction, can be received at the server from the device.
  • the sensors 162 may include a number of sensors configured to sense information about an environment in which the vehicle 158 is located, as well as one or more actuators configured to modify a position and/or orientation of the sensors.
  • the sensors can include a global positioning system ("GPS"), an inertial measurement unit (“IMU”), a RADAR unit, a laser rangefinder and/or one or more LIDAR units, and/or a camera.
  • GPS global positioning system
  • IMU inertial measurement unit
  • RADAR unit a RADAR unit
  • LIDAR units laser rangefinder and/or one or more LIDAR units
  • the sensors 162 may be implemented as multiple sensor units each mounted to the vehicle in a respective position (e.g., top side, bottom side, front side, back side, right side, left side, etc.). Other sensors are possible as well.
  • the vehicle communications system 164 may be any system communicatively coupled (via wires or wirelessly) to one or more other vehicles, sensors, or other entities, either directly and/or via a communications network.
  • the wireless communication system 164 may include an antenna and a chipset for communicating with the other vehicles, sensors, servers, and/or other entities either directly or via a communications network.
  • the chipset or wireless communication system 164 in general may be arranged to communicate according to one or more types of wireless communication (e.g., protocols) such as BLUETOOTH, communication protocols described in IEEE 802.11 (including any IEEE 802.11 revisions), cellular technology (such as V2X, V2V, GSM, CDMA, UMTS, EV-DO, WiMAX, or LTE), ZIGBEE, dedicated short range communications (DSRC), and radio frequency identification (“RFID”) communications, among other possibilities.
  • the wireless communication system 164 may take other forms as well.
  • the propulsion system 166 may be configured to provide powered motion for the vehicle 158.
  • the propulsion system 166 may include various components to provide such motion, including an engine/motor, an energy source, a transmission, and wheels/tires.
  • the engine/motor may include any combination of an internal combustion engine, an electric motor (that can be powered by an electrical battery, fuel cell, and/or other energy storage device), and/or a steam engine. Other motors and engines are possible as well.
  • the control system 168 may be configured to control operation of the vehicle 158 and its components.
  • the control system 168 may include various components, including a steering unit, a throttle, a brake unit, a perception system, a navigation or pathing system, and an obstacle avoidance system.
  • a perception system may be any system configured to process and analyze images and/or sensor data captured by the sensors (e.g., a camera, RADAR and/or LIDAR) of the vehicle 158 in order to identify objects and/or features in the environment in which the vehicle 158 is located, including, for example, traffic signals and obstacles.
  • the perception system may use an object recognition algorithm, a Structure from Motion ("SFM") algorithm, video tracking, or other computer vision techniques.
  • SFM Structure from Motion
  • the perception system may additionally be configured to map the environment, track objects, estimate the speed of objects, etc.
  • the overall system can comprise a perception subsystem for identifying objects, a planning subsystem for planning a smooth driving path around the obstacles, and a control subsystem for executing the path from the planner.
  • the power supply 170 may be a source of energy that powers the engine/motor of the vehicle 158 in full or in part and/or powers the electrical equipment of the vehicle 158.
  • the engine/motor of the vehicle may be configured to convert the power supply 170 into mechanical energy.
  • Examples of energy sources for the power supply 170 include gasoline, diesel, propane, other compressed gas-based fuels, ethanol, solar panels, batteries, and other sources of electrical power.
  • the energy source(s) may additionally or alternatively include any combination of fuel tanks, batteries, capacitors, and/or flywheels. In some embodiments, the energy source may provide energy for other systems of the vehicle 158 as well.
  • the power supply 170 is an electric battery that is communicatively coupled to the BMS 174.
  • the BMS 174 monitors the various characteristics of the power supply 170, including battery temperature, battery voltage, battery current, and battery charging and discharging data. This information can be stored locally by the BMS 174 and/or the computing system 160.
  • the BMS 174 can also transmit such monitored information via the vehicle communications system 164 to an external storage device (e.g., in the cloud).
  • the BMS 174 may regulate the operating conditions of the power supply 170, e.g., cooling the battery temperature to within a predefined threshold temperature.
  • the computing system 160 can be configured to generate a SoH using a supervised DNN of the present disclosure.
  • a computing system comprises: at least one data storage configured to store computer program instructions; and at least one processor communicatively coupled to the at least one data storage, the at least one processor is configured to execute the computer program instructions to perform the following, comprising: determining a state of health indication for a battery, further comprising the steps of: receiving characteristics and process parameters for the battery; training a first supervised deep neural network, where the received characteristics are inputs to the first supervised DNN and the received process parameters are outputs to the supervised DNN, training an unsupervised AI estimator based on extracted features from the first supervised DNN, where the received characteristics are inputs to the unsupervised AI estimator; training a second supervised DNN using identified clusters from the unsupervised AI estimator; validating the identified clusters with state of health indications; and determining a state of health indication for the battery using the second supervised DNN.
  • the at least one data storage can be a non-transitory computer readable medium encoded with instructions that when executed by the at least one processor causes the processor to carry out the instructions.
  • the SoH indication can be inputted to the BMS 174 for the BMS 174 to adjust the operating conditions of the power supply 170.
  • the supervised DNN for generating a SoH indication can be performed by one or more processors of the BMS 174.
  • the user interface system 172 may include software, a human-machine interface ("HMI"), and/or peripherals that are configured to allow the vehicle 158 to interact with external sensors, other vehicles, external computing devices, and/or a user.
  • the peripherals may include, for example, a wireless communication system, a touchscreen, a microphone, and/or a speaker.
  • SoH indication or related metric e.g., SoC
  • SoC SoC
  • the vehicle 158 may include one or more elements in addition to or instead of those shown.
  • the vehicle 158 may include one or more additional interfaces and/or power supplies.
  • Other additional components are possible as well.
  • the data storage of the computing system 160 may further include instructions executable by the processor of the computing system 160 to control and/or communicate with the additional components.
  • one or more components or systems may be removably mounted on or otherwise connected (mechanically or electrically) to the vehicle 158 using wired or wireless connections.
  • a cloud service and/or backend server can be configured to perform DNN compression (e.g., using similarity-based filter pruning, quantization, or other DNN compression method).
  • DNN compression e.g., using similarity-based filter pruning, quantization, or other DNN compression method.
  • the cloud service and/or backend server can deploy the DNN to the vehicle 158 and the vehicle can perform inference using the compressed DNN on embedded hardware of the vehicle 158, e.g., by the computing system 160.
  • the supervised DNN for generating a SoH indication for the power supply 170 can be generated in real-time in the vehicle 158.
  • the computing system 160 can run the DNN predictions at runtime on embedded hardware that may have limited computing capabilities. Thus, multiple functions can be run simultaneously on the computing system.
  • the compressed DNN size can lead to a small footprint in memory of the computing system 160 and can be transmitted quickly over wireless connections. Thus, when an improved DNN version is released, the improved DNN can be easily deployed to the vehicle 158 via the vehicle communications system 164 and processed by the computing system 160.
  • disclosed embodiments use software for performing functionality to enable measurement and analysis of data, at least in part, using software code stored on one or more non-transitory computer readable mediums running on one or more processors in a transportation vehicle.
  • Such software and processors may be combined to constitute at least one controller coupled to other components of the transportation vehicle to support and provide autonomous and/or assistive transportation vehicle functionality in conjunction with vehicle navigation systems, and multiple sensors.
  • Such components may be coupled with the at least one controller for communication and control via a CAN bus of the transportation vehicle or other busses (e.g., Flexray).
  • Disclosed embodiments include the methods described herein and their equivalents, non-transitory computer readable media programmed to carry out the methods and a computing system configured to carry out the methods. Further, included is a vehicle comprising components that include any of the methods, non-transitory computer readable media programmed to implement the instructions or carry out the methods, and systems to carry out the methods.
  • the computing system, and any sub-computing systems will typically include a machine-readable storage medium containing executable code; one or more processors; memory coupled to the one or more processors; an input device, and an output device connected to the one or more processors to execute the code.
  • a machine-readable medium may include any mechanism for storing or transmitting information in a form readable by a machine, such as a computer processor. The information may be stored, for example, in volatile or non-volatile memory. Additionally, embodiment functionality may be implemented using embedded devices and online connection to cloud computing infrastructure available through radio connection (e.g., wireless communication) with such infrastructure.
  • embodiments of the subject matter and the functional operations described in this specification can be implemented in digital electronic circuitry, in tangibly-embodied computer software or firmware, in computer hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of one or more of them.
  • Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible non-transitory storage medium for execution by, or to control the operation of, data processing apparatus.
  • the computer storage medium can be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of them.
  • the program instructions can be encoded on an artificially-generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus.
  • an artificially-generated propagated signal e.g., a machine-generated electrical, optical, or electromagnetic signal

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Secondary Cells (AREA)
  • Charge And Discharge Circuits For Batteries Or The Like (AREA)

Claims (15)

  1. Un système informatique comprenant :
    au moins une mémoire de données configurée pour stocker des instructions de programme informatique ; et
    au moins un processeur couplé sur le plan de la communication à l'au moins une
    mémoire de données,
    l'au moins un processeur étant configuré pour exécuter les instructions du programme informatique afin de réaliser les étapes suivantes, à savoir :
    la détermination d'une indication de l'état de santé d'une batterie, comprenant en outre les étapes suivantes :
    la réception (40) des caractéristiques et des paramètres de processus de la batterie ;
    l'entraînement (42) d'un premier réseau neuronal profond supervisé, DNN, les caractéristiques reçues étant des entrées dans le premier DNN supervisé et les paramètres de processus reçus étant des sorties du premier DNN supervisé ;
    l'entraînement (44) d'un estimateur d'intelligence artificielle (IA) non supervisé à l'aide d'un ou plusieurs procédés de regroupement basés sur les caractéristiques extraites du premier DNN supervisé, les caractéristiques reçues étant des entrées dans l'estimateur d'IA non supervisé ;
    l'entraînement (46) d'un second DNN supervisé à l'aide de groupes identifiés par l'estimateur d'IA non supervisé, les caractéristiques reçues étant des entrées dans le second DNN supervisé ;
    la validation (48) des groupes identifiés avec des indications de l'état de santé ; et
    la détermination (50) de l'indication de l'état de santé de la batterie à l'aide du second DNN supervisé.
  2. Le système informatique selon la revendication 1, dans lequel, après la troisième étape d'entraînement et avant l'étape de validation, la répétition de l'étape d'entraînement de l'estimateur d'IA non supervisé et de l'étape d'entraînement du second DNN supervisé est réalisée jusqu'à ce qu'une condition de convergence soit satisfaite.
  3. Le système informatique selon la revendication 1, dans lequel, lors de la deuxième étape d'entraînement, un ou plusieurs procédés de regroupement comprennent la division des valeurs continues des sorties de l'estimateur d'IA non supervisé en cases discrétisées.
  4. Le système informatique selon la revendication 1, dans lequel, lors de la deuxième étape d'entraînement, un ou plusieurs procédés de regroupement comprennent l'application d'un DNN de régression aux sorties de l'estimateur d'IA non supervisé.
  5. Le système informatique selon la revendication 1, dans lequel, lors de la deuxième étape d'entraînement, le ou les procédés de regroupement comprennent l'application du regroupement de sous-espaces avec un auto-encodeur pour la réduction de dimension aux sorties de l'estimateur d'IA non supervisé.
  6. Le système informatique selon la revendication 1, dans lequel les caractéristiques reçues incluent au moins une caractéristique d'un groupe comprenant la température de la batterie, la tension de la batterie et le courant électrique de la batterie.
  7. Le système informatique selon la revendication 1, dans lequel les paramètres de processus reçus incluent au moins l'un des groupes comprenant le taux de charge de la batterie, l'état de charge de la batterie et la température du processus.
  8. Procédé informatisé permettant de déterminer une indication de l'état de santé d'une batterie, comprenant:
    la réception (40) des caractéristiques et des paramètres de processus de la batterie ;
    l'entraînement (42) d'un premier réseau neuronal profond supervisé, DNN, les caractéristiques reçues étant des entrées dans le premier DNN supervisé et les paramètres de processus reçus étant des sorties du premier DNN supervisé ;
    l'entraînement (44) d'un estimateur d'intelligence artificielle (IA) non supervisé à l'aide d'un ou plusieurs procédés de regroupement basés sur les caractéristiques extraites du premier DNN supervisé, les caractéristiques reçues étant des entrées dans l'estimateur d'IA non supervisé ;
    l'entraînement (46) d'un second DNN supervisé à l'aide de groupes identifiés par l'estimateur d'IA non supervisé, les caractéristiques reçues étant des entrées dans le second DNN supervisé ;
    la validation (48) des groupes identifiés avec des indications de l'état de santé ; et
    la détermination (50) de l'indication de l'état de santé de la batterie à l'aide du second DNN supervisé.
  9. Le procédé informatisé selon la revendication 8, dans lequel, après la troisième étape d'entraînement et avant l'étape de validation, la répétition de l'étape d'entraînement de l'estimateur d'IA non supervisé et de l'étape d'entraînement du second DNN supervisé est réalisée.
  10. Le procédé informatisé selon la revendication 8, dans lequel, lors de la deuxième étape d'entraînement, un ou plusieurs procédés de regroupement comprennent la division des valeurs continues des sorties de l'estimateur d'IA non supervisé en cases discrétisées.
  11. Le procédé informatisé selon la revendication 8, dans lequel, lors de la deuxième étape d'entraînement, un ou plusieurs procédés de regroupement comprennent l'application d'un DNN de régression aux sorties de l'estimateur d'IA non supervisé.
  12. Le procédé informatisé selon la revendication 8, dans lequel, lors de la deuxième étape d'entraînement, le ou les procédés de regroupement comprennent l'application du regroupement de sous-espaces avec un auto-encodeur pour la réduction de dimension aux sorties de l'estimateur d'IA non supervisé.
  13. Le procédé informatisé selon la revendication 8, dans lequel les caractéristiques reçues incluent au moins une caractéristique d'un groupe comprenant la température de la batterie, la tension de la batterie et le courant électrique de la batterie.
  14. Le procédé informatisé selon la revendication 8, dans lequel les paramètres de processus reçus incluent au moins l'un des groupes comprenant le taux de charge de la batterie, l'état de charge de la batterie et la température du processus.
  15. Support informatique non transitoire codé avec des instructions qui, lorsqu'elles sont exécutées par au moins un processeur, amènent ce dernier à effectuer les opérations suivantes :
    la réception (40) des caractéristiques et des paramètres de processus d'une batterie ;
    l'entraînement (42) d'un premier réseau neuronal profond supervisé, DNN, les caractéristiques reçues étant des entrées dans le premier DNN supervisé et les paramètres de processus reçus étant des sorties du premier DNN supervisé ;
    l'entraînement (44) d'un estimateur d'intelligence artificielle (IA) non supervisé à l'aide d'un ou plusieurs procédés de regroupement basés sur les caractéristiques extraites du premier DNN supervisé, les caractéristiques reçues étant des entrées dans l'estimateur d'IA non supervisé ;
    l'entraînement (46) d'un second DNN supervisé à l'aide de groupes identifiés par l'estimateur d'IA non supervisé, les caractéristiques reçues étant des entrées dans le second DNN supervisé ;
    la validation (48) des groupes identifiés avec des indications de l'état de santé ; et
    la détermination (50) d'une indication de l'état de santé de la batterie à l'aide du second DNN supervisé.
EP21766146.1A 2020-09-03 2021-08-19 Estimateur d'état de santé de batterie Active EP4208725B1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US17/011,103 US11360155B2 (en) 2020-09-03 2020-09-03 Battery state of health estimator
PCT/EP2021/073081 WO2022048920A1 (fr) 2020-09-03 2021-08-19 Estimateur d'état de santé de batterie

Publications (2)

Publication Number Publication Date
EP4208725A1 EP4208725A1 (fr) 2023-07-12
EP4208725B1 true EP4208725B1 (fr) 2024-03-27

Family

ID=77655540

Family Applications (1)

Application Number Title Priority Date Filing Date
EP21766146.1A Active EP4208725B1 (fr) 2020-09-03 2021-08-19 Estimateur d'état de santé de batterie

Country Status (4)

Country Link
US (1) US11360155B2 (fr)
EP (1) EP4208725B1 (fr)
CN (1) CN116034282A (fr)
WO (1) WO2022048920A1 (fr)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102424916B1 (ko) * 2021-10-14 2022-07-22 모나일렉트릭 주식회사 배터리 진단 방법 및 그 장치
US11921163B2 (en) * 2021-12-16 2024-03-05 Volkswagen Aktiengesellschaft Modular machine learning structure for electric vehicle battery state of health estimation
US20240062514A1 (en) * 2022-08-19 2024-02-22 X Development Llc Diversity quantification
US11707987B1 (en) * 2022-12-06 2023-07-25 Mercedes-Benz Group AG Vehicle simulating method and system
EP4390421A1 (fr) * 2022-12-22 2024-06-26 Ningbo Geely Automobile Research & Development Co., Ltd. Procédé d'estimation de l'état de santé d'une batterie
CN117556715B (zh) * 2024-01-12 2024-03-26 湖南大学 基于信息融合的典型环境下智能电表退化分析方法及系统

Family Cites Families (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108604310B (zh) * 2015-12-31 2022-07-26 威拓股份有限公司 用于使用神经网络架构来控制分配系统的方法、控制器和系统
WO2019017991A1 (fr) * 2017-07-21 2019-01-24 Quantumscape Corporation Modèle prédictif pour estimer des états de batterie
US11147459B2 (en) * 2018-01-05 2021-10-19 CareBand Inc. Wearable electronic device and system for tracking location and identifying changes in salient indicators of patient health
US11429862B2 (en) * 2018-03-20 2022-08-30 Sri International Dynamic adaptation of deep neural networks
US11080590B2 (en) * 2018-03-21 2021-08-03 Nvidia Corporation Stereo depth estimation using deep neural networks
EP3620983B1 (fr) * 2018-09-05 2023-10-25 Sartorius Stedim Data Analytics AB Procédé mis en uvre par ordinateur, produit de programme informatique et système d'analyse de données
KR102206181B1 (ko) * 2018-12-19 2021-01-22 엘지전자 주식회사 단말기 및 그의 동작 방법
KR102137151B1 (ko) * 2018-12-27 2020-07-24 엘지전자 주식회사 노이즈 캔슬링 장치 및 그 방법
US11325494B2 (en) * 2019-02-25 2022-05-10 Toyota Research Institute, Inc. Systems, methods, and storage media for determining a target battery charging level for a drive route
US11084387B2 (en) * 2019-02-25 2021-08-10 Toyota Research Institute, Inc. Systems, methods, and storage media for arranging a plurality of cells in a vehicle battery pack
US11555858B2 (en) * 2019-02-25 2023-01-17 Toyota Research Institute, Inc. Systems, methods, and storage media for predicting a discharge profile of a battery pack
US11065978B2 (en) * 2019-02-25 2021-07-20 Toyota Research Institute, Inc. Systems, methods, and storage media for adapting machine learning models for optimizing performance of a battery pack
KR102281590B1 (ko) * 2019-07-31 2021-07-29 엘지전자 주식회사 음성인식 성능 향상을 위한 비 지도 가중치 적용 학습 시스템 및 방법, 그리고 기록 매체
US11562236B2 (en) * 2019-08-20 2023-01-24 Lg Electronics Inc. Automatically labeling capability for training and validation data for machine learning
US10877634B1 (en) * 2019-10-03 2020-12-29 Raytheon Company Computer architecture for resource allocation for course of action activities
CN111027625B (zh) 2019-12-10 2023-08-08 燕山大学 基于SAE和K-means聚类算法的电池筛选方法

Also Published As

Publication number Publication date
US11360155B2 (en) 2022-06-14
CN116034282A (zh) 2023-04-28
WO2022048920A1 (fr) 2022-03-10
EP4208725A1 (fr) 2023-07-12
US20220065939A1 (en) 2022-03-03

Similar Documents

Publication Publication Date Title
EP4208725B1 (fr) Estimateur d'état de santé de batterie
You et al. Real-time state-of-health estimation for electric vehicle batteries: A data-driven approach
KR102241683B1 (ko) 배터리의 상태를 추정하는 방법 및 장치
Modi et al. Estimation of energy consumption of electric vehicles using deep convolutional neural network to reduce driver’s range anxiety
US10310022B2 (en) Battery state estimation apparatus and method
JP2020145186A (ja) 電池パックの性能を最適化するための機械学習モデルを適用するためのシステム、方法および記憶媒体
Mamo et al. Long short-term memory with attention mechanism for state of charge estimation of lithium-ion batteries
US11551094B2 (en) System and method for deep neural network compression
Zhao et al. Battery prognostics and health management from a machine learning perspective
Modi et al. Convolutional neural network–bagged decision tree: a hybrid approach to reduce electric vehicle’s driver’s range anxiety by estimating energy consumption in real-time
WO2022133090A1 (fr) Génération et évaluation adaptatives de scénarios critiques de véhicule autonome
Eaty et al. Digital twin for electric vehicle battery management with incremental learning
Beltran et al. Machine learning techniques suitability to estimate the retained capacity in lithium-ion batteries from partial charge/discharge curves
Pau et al. Microcontroller architectures for battery state of charge prediction with tiny neural networks
Nascimento et al. Li-ion battery aging with hybrid physics-informed neural networks and fleet-wide data
US11614491B2 (en) Systems and methods for predicting the cycle life of cycling protocols
US20240210477A1 (en) Battery management system using graph neural network
El Marghichi Estimation of battery capacity using the enhanced self-organization maps
Liu et al. A denoising svr-mlp method for remaining useful life prediction of lithium-ion battery
Bockrath et al. Generalized State of Health Estimation Approach based on Neural Networks for Various Lithium-Ion Battery Chemistries
Di Fiore et al. Rapid Assessment of Incipient Multimodal Faults of Complex Aerospace Systems
Corbetta et al. Hybrid Modeling of Unmanned Aerial Vehicle Electric Powertrain for Fault Detection and Diagnostics
US20240217388A1 (en) Probabilistic modelling of electric vehicle charging and driving usage behavior with hidden markov model-based clustering
Saba et al. Advancing Electric Vehicle Battery Analysis With Digital Twins in Intelligent Transportation Systems
Elouazzani et al. Advancing Lithium-Ion Battery Management with Deep Learning: A Comprehensive Review

Legal Events

Date Code Title Description
STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: UNKNOWN

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE

PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE

17P Request for examination filed

Effective date: 20230321

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

GRAP Despatch of communication of intention to grant a patent

Free format text: ORIGINAL CODE: EPIDOSNIGR1

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: GRANT OF PATENT IS INTENDED

DAV Request for validation of the european patent (deleted)
DAX Request for extension of the european patent (deleted)
INTG Intention to grant announced

Effective date: 20231121

GRAS Grant fee paid

Free format text: ORIGINAL CODE: EPIDOSNIGR3

GRAA (expected) grant

Free format text: ORIGINAL CODE: 0009210

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE PATENT HAS BEEN GRANTED

AK Designated contracting states

Kind code of ref document: B1

Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR

REG Reference to a national code

Ref country code: GB

Ref legal event code: FG4D

REG Reference to a national code

Ref country code: CH

Ref legal event code: EP

REG Reference to a national code

Ref country code: DE

Ref legal event code: R096

Ref document number: 602021011041

Country of ref document: DE

REG Reference to a national code

Ref country code: IE

Ref legal event code: FG4D

P01 Opt-out of the competence of the unified patent court (upc) registered

Effective date: 20240502